Implied Entity Reasoning In Text-to-scene Conversion

Human beings perceive the objective world through their facial features,while it is more intuitive and convenient to acquire the information through visual sense than to understand the textual content.However,with the advent of the information age,the information stored in the form of text presents the massive growth trend,and people’s demand for information visualization is growing.The text-to-scene conversion is to make the information stored in the form of text displayed in a visual way such as 3D scene or animation in the computer.However,text-to-scene conversion is still in the stage of development,the relevant theoretical and technical research is not mature.This paper focuses on information visualization extraction and simple scene entity in the text-to-scene conversion,the main work is as follows:1.Introduces the existing text-to-scene conversion system and provide the logic framework of the Text-to-scene conversion.Analyze the related relevant natural language processing method and the relevant scene generated technologies.2.For the extraction problem of information visualization in the text-to-scene conversion,based on the constructing the experimental data set,this thesis finds a suitable tool for Chinese natural language processing through horizontal comparison methods.Then on the basis of natural language processing,this thesis puts forward the extraction rules of visual scene information,and realizes the recognition from text to scene information from the point of morphology and syntax.3.For the simple scene entity problem,this thesis makes semantic reasoning for input information by Hidden Markov Model,and gets the implicit entities in the input text based on the extraction rules of visual scene information,finally displays in the generated visual for scene enhancement.4.The extraction rules of scene information and scene information enhancement methods mentioned in this thesis are implemented and verified by the combination of 2D background and 3D solid model through the establishment of spatial knowledge base and entity model library.The result showed that the automatic enhancement function proposed in this thesis can realize the extraction of text information visualization and scene automatic enhancement,improve the automatic conversion efficiency of text-to-scene conversion and performance in a certain degree.